1. Field of the Invention
This invention relates to peer-to-peer networking, and more particularly to a system and method for establishing reliable connections between peers in a peer-to-peer networking environment.
2. Description of the Related Art
The Internet has three valuable fundamental assets—information, bandwidth, and computing resources—all of which are vastly underutilized, partly due to the traditional client-server computing model. No single search engine or portal can locate and catalog the ever-increasing amount of information on the Web in a timely way. Moreover, a huge amount of information is transient and not subject to capture by techniques such as Web crawling. For example, research has estimated that the world produces two exabytes or about 2×1018 bytes of information every year, but only publishes about 300 terabytes or about 3×1012 bytes. In other words, for every megabyte of information produced, only one byte gets published. Moreover, Google claims that it searches about only 1.3×10^8 web pages. Thus, finding useful information in real time is increasingly difficult.
Although miles of new fiber have been installed, the new bandwidth gets little use if everyone goes to one site for content and to another site for auctions. Instead, hot spots just get hotter while cold pipes remain cold. This is partly why most people still feel the congestion over the Internet while a single fiber's bandwidth has increased by a factor of 10^6 since 1975, doubling every 16 months.
New processors and storage devices continue to break records in speed and capacity, supporting more powerful end devices throughout the network. However, computation continues to accumulate around data centers, which have to increase their workloads at a crippling pace, thus putting immense pressure on space and power consumption.
Finally, computer users in general are accustomed to computer systems that are deterministic and synchronous in nature, and think of such a structure as the norm. For example, when a browser issues a URL request for a Web page, the output is typically expected to appear shortly afterwards. It is also typically expected that everyone around the world will be able to retrieve the same page from the same Web server using the same URL.
The term peer-to-peer networking or computing (often referred to as P2P) may be applied to a wide range of technologies that greatly increase the utilization of information, bandwidth, and computing resources in the Internet. Frequently, these P2P technologies adopt a network-based computing style that neither excludes nor inherently depends on centralized control points. Apart from improving the performance of information discovery, content delivery, and information processing, such a style also can enhance the overall reliability and fault-tolerance of computing systems.
Peer-to-peer (P2P) computing, embodied by applications like Napster, Gnutella, and Freenet, has offered a compelling and intuitive way for Internet users to find and share resources directly with each other, often without requiring a central authority or server. As much as these diverse applications have broken new ground, they typically address only a single function, run primarily only on a single platform, and are unable to directly share data with other, similar applications.
Many peer-to-peer systems are built for delivering a single type of service. For example, Napster provides music file sharing, Gnutella provides generic file sharing, and AIM provides instant messaging. Given the diverse characteristics of these services and the lack of a common underlying P2P infrastructure, each P2P software vendor tends to create incompatible systems—none of them able to interoperate with one another. This means each vendor creates its own P2P user community, duplicating efforts in creating software and system primitives commonly used by all P2P systems. Moreover, for a peer to participate in multiple communities organized by different P2P implementations, the peer must support multiple implementations, each for a distinct P2P system or community, and serve as the aggregation point.
Many P2P systems today offer their features or services through a set of APIs that are delivered on a particular operating system using a specific networking protocol. For example, one system might offer a set of C++ APIs, with the system initially running only on Windows, over TCP/IP, while another system offers a combination and C and Java APIs, running on a variety of UNIX systems, over TCP/IP but also requiring HTTP. A P2P developer is then forced to choose which set of APIs to program to, and consequently, which set of P2P customers to target. Because there is little hope that the two systems will interoperate, if the developer wants to offer the same service to both communities, they have to develop the same service twice for two P2P platforms or develop a bridge system between them. Both approaches are inefficient and impractical considering the dozens of P2P platforms in existence.
Many P2P systems, especially those being offered by upstart companies, tend to choose one operating system as their target deployment platform. The cited reason for this choice is to target the largest installed base and the fastest path to profit. The inevitable result is that many dependencies on platform-specific features are designed into (or just creep into) the system. This is often not the consequence of technical desire but of engineering reality with its tight schedules and limited resources.
This approach is clearly shortsighted. Even though the earliest demonstration of P2P capabilities are on platforms in the middle of the computing hardware spectrum, it is very likely that the greatest proliferation of P2P technology will occur at the two ends of the spectrum—large systems in the enterprise and consumer-oriented small systems. In fact, betting on any particular segment of the hardware or software system is not future proof.
FIGS. 1A and 1B are examples illustrating the peer-to-peer model. FIG. 1A shows two peer devices 104A and 104B that are currently connected. Either of the two peer devices 104 may serve as a client of or a server to the other device. FIG. 1B shows several peer devices 104 connected over the network 106 in a peer group. In the peer group, any of the peer devices 104 may serve as a client of or a server to any of the other devices.
Prior art peer-to-peer systems are generally built for delivering a single type of service, for example a music file sharing service, a generic file sharing service, or an instant messaging service. Given the diverse characteristics of these services and given the lack of a common underlying peer-to-peer infrastructure, each vendor tends to form various peer-to-peer “silos”. In other words, the prior art peer-to-peer systems typically do not interoperate with each other. This means each vendor has to create its own peer-to-peer user community, duplicating efforts in creating primitives commonly used by peer-to-peer systems such as peer discovery and peer communication.
Discovery in a peer-to-peer environment may be based on centralized discovery with a centralized index. This method is used by such peer-to-peer applications as Napster and AIM. Discovery based on a centralized index may be efficient, deterministic, and well suited for a static environment. Such a method of discovery may also provide centralized control, provide a central point of failure, and provide easy denial of services. However, such a method of discovery may be expensive to scale and may degrade with aging.
Discovery in a peer-to-peer environment may also be based on net crawling. This method is used by such peer-to-peer applications as Gnutella and FreeNet. Discovery based on net crawling may be simple, adaptive, deterministic, inexpensive to scale, well suited for a dynamic environment, and may be difficult to attack. Such a method of discovery may also improve with aging. However, such a method of discovery may provide slower discovery than centralized control.
In a peer-to-peer environment, assume there is a peer-to-peer community offering a search capability for its members, where one member can post a query and other members can hear and respond to the query. One member is a Napster user and has implemented a feature so that, whenever a query is received seeking an MP3 file, this member will look up the Napster directory and then respond to the query with information returned by the Napster system. Here, a member without any knowledge of Napster may benefit because another member implemented a bridge to connect their peer-to-peer system to Napster. This type of bridging is very useful, but when the number of services is large, pair-wise bridging becomes more difficult and undesirable. Thus, it may be desirable to provide a platform bridge that may be used to connect various peer-to-peer systems together.
In another example, one engineering group requires a sizable storage capability, but also with redundancy to protect data from sudden loss. A common solution is to purchase a storage system with a large capacity and mirrored disks. Another engineering group later decides to purchase the same system. Both groups end up with a lot of extra capacity, and have to pay higher prices for the mirroring feature. Thus, it may be desirable to provide a mechanism by which each group may buy a simple storage system without the mirroring feature, where the disks can then discover each other automatically, form a storage peer group, and offer mirroring facilities using their spare capacity.
As yet another example, many devices such as cell phones, pagers, wireless email devices, Personal Digital Assistants (PDAs), and Personal Computers (PCs) may carry directory and calendar information. Currently, synchronization among the directory and calendar information on these devices is very tedious, if not impossible. Often, a PC becomes the central synchronization point, where every other device has to figure out a way to connect to the PC (using serial port, parallel port, IrDA, or other method) and the PC must have the device driver for every device that wishes to connect. Thus, it may be desirable to provide a mechanism by which these devices may interact with each other, without extra networking interfaces except those needed by the devices themselves, utilizing a common layer of communication and data exchange.
Reliable delivery of data on a network may include the notion of guaranteed delivery (barring a connection drop or major network, hardware, and/or software failure) and ordering of the data. Some connections between peers in a peer-to-peer environment may provide unidirectional, asynchronous delivery of data. Such connections, however, typically do not provide reliable delivery of the data. Some connections between peers in a peer-to-peer environment may be implemented over network connections using a network protocol that provides reliable delivery of data packets (e.g. TCP/IP). However, such network protocols typically only provides reliable delivery on a direct connection between two nodes; if there are one or more hops on the network connection between the two peer nodes, then delivery of data may not be reliable. In addition, connections between peers may pass through one or more firewalls and/or gateways, further hindering reliable delivery of data.
While the invention is described herein by way of example for several embodiments and illustrative drawings, those skilled in the art will recognize that the invention is not limited to the embodiments or drawings described. It should be understood, that the drawings and detailed description thereto are not intended to limit the invention to the particular form disclosed, but on the contrary, the intention is to cover all modifications, equivalents and alternatives falling within the spirit and scope of the present invention as defined by the appended claims. The headings used herein are for organizational purposes only and are not meant to be used to limit the scope of the description or the claims. As used throughout this application, the word “may” is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words “include”, “including”, and “includes” mean including, but not limited to.